Zhijie Sun
ABSTRACT
In the era of self-media, the spread of fake news is more widespread and rapid because anyone can become an editor of the news. It leads to a large number of short news texts. However, the unchecked dissemination and sharing of news information have led to a continuous emergence of fake news events. It not only misleads readers but also has a detrimental impact on society. Besides, short news texts are semantically sparse and need more contextual solid connections. It is difficult to extract text features and achieve high error detection efficiency. Therefore, this paper has proposed a heterogeneous graph attention network that includes multiple text-related features. The network breaks the traditional way of only connecting text features in graph neural networks, extracts various external knowledge and text feature information from the news to construct the graph network, and establishes connections for different text features to enhance semantic understanding. Additionally, connecting external knowledge bases is helpful in eliminating news entity word ambiguities. Then, the heterogeneous graph is embedded into a dual-attention mechanism at both node and pattern levels, capturing the importance of different adjacent nodes, reducing the weight of noisy nodes, and accurately identifying valid information. Experimental results demonstrate that the proposed method in this paper outperforms better than several baseline models, such as TextGCN in terms of accuracy.
- Vosoughi S, Roy D, Aral S. The spread of true and false news online[J]. science, 2018, 359(6380): 1146-1151.Google Scholar
- Zhou X, Zafarani R, Shu K, Fake news: Fundamental theories, detection strategies and challenges[C]//Proceedings of the twelfth ACM international conference on web search and data mining. 2019: 836-837.Google Scholar
- Dai Z, Yang Z, Yang Y, Transformer-XL: Attentive Language Models beyond a Fixed-Length Context[C]//Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics. 2019: 2978-2988.Google Scholar
- Kenton J D M W C, Toutanova L K. Bert: Pre-training of deep bidirectional transformers for language understanding[C]//Proceedings of naacL-HLT. 2019, 1: 2.Google Scholar
- Zhang Z, Han X, Liu Z, ERNIE: Enhanced Language Representation with Informative Entities[C]//Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics. 2019: 1441-1451.Google Scholar
- Ren F, She T. Utilizing external knowledge to enhance semantics in emotion detection in conversation[J]. IEEE Access, 2021, 9: 154947-154956.Google ScholarCross Ref
- Zhuang L, Wayne L, Ya S, A robustly optimized BERT pre-training approach with post-training[C]//Proceedings of the 20th chinese national conference on computational linguistics. 2021: 1218-1227.Google Scholar
- Tenney I, Das D, Pavlick E. BERT Rediscovers the Classical NLP Pipeline[C]//Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics. 2019: 4593-4601.Google Scholar
- Wang S, Guo Y, Wang Y, SMILES-BERT: large scale unsupervised pre-training for molecular property prediction[C]//Proceedings of the 10th ACM international conference on bioinformatics, computational biology and health informatics. 2019: 429-436.Google Scholar
- Nguyen T P, Razniewski S, Weikum G. Advanced semantics for commonsense knowledge extraction[C]//Proceedings of the Web Conference 2021. 2021: 2636-2647.Google Scholar
- Vaswani A, Shazeer N, Parmar N, Attention is all you need[J]. Advances in neural information processing systems, 2017, 30.Google Scholar
- Peters M E, Neumann M, Iyyer M, Deep contextualized word representations[J]// Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long Papers).2018: 2227–2237.Google Scholar
- Radford A, Narasimhan K, Salimans T, Improving language understanding by generative pre-training[J]. 2018.Google Scholar
- Velickovic P, Cucurull G, Casanova A, Graph attention networks[J]. stat, 2017, 1050(20): 10.48550.Google Scholar
- Yao L, Mao C, Luo Y. Graph convolutional networks for text classification[C]//Proceedings of the AAAI conference on artificial intelligence. 2019, 33(01): 7370-7377.Google Scholar
- Zhang Y, Yu X, Cui Z, Every Document Owns Its Structure: Inductive Text Classification via Graph Neural Networks[C]//Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics. 2020: 334-339.Google Scholar
- Ding K, Wang J, Li J, Be more with less: Hypergraph attention networks for inductive text classification[C]//2020 Conference on Empirical Methods in Natural Language Processing, EMNLP 2020. Association for Computational Linguistics (ACL), 2020: 4927-4936.Google Scholar
- Linmei H, Yang T, Shi C, Heterogeneous graph attention networks for semi-supervised short text classification[C]//Proceedings of the 2019 conference on empirical methods in natural language processing and the 9th international joint conference on natural language processing (EMNLP-IJCNLP). 2019: 4821-4830.Google Scholar
- Ren Y, Zhang J. Fake news detection on news-oriented heterogeneous information networks through hierarchical graph attention[C]//2021 International Joint Conference on Neural Networks (IJCNN). IEEE, 2021: 1-8.Google Scholar
- Mehta N, Pacheco M L, Goldwasser D. Tackling Fake News Detection by Continually Improving Social Context Representations using Graph Neural Networks[C]//Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2022: 1363-1380.Google Scholar
Index Terms
- Graph Attention Network for Short Text Type News
Recommendations
Planar graph colorings without short monochromatic cycles
It is well known that every planar graph G is 2-colorable in such a way that no 3-cycle of G is monochromatic. In this paper, we prove that G has a 2-coloring such that no cycle of length 3 or 4 is monochromatic. The complete graph K5 does not admit ...
Equitable colorings of planar graphs without short cycles
An equitable coloring of a graph is a proper vertex coloring such that the sizes of every two color classes differ by at most 1. Chen, Lih, and Wu conjectured that every connected graph G with maximum degree @D>=2 has an equitable coloring with @D ...
Heterogeneous graph convolutional neural network for short text classification
Most existing short text classification methods treat each phrase as an independent homogeneous distribution, thus losing the association information between sentences. To solve this problem, we propose a heterogeneous GCN for short text classification. ...
Comments